Market structure analysis

Table 2.4: Coal quality by exporter

kcal/kg GJ/t κx in t/GJ

Australia 6400 26.80 0.03732

Indonesia 5450 22.82 0.04382

South Africa 6260 26.21 0.03815

Russia West 6400 26.80 0.03732

Russia East 6300 26.38 0.03791

China 6200 25.96 0.03852

Colombia 6375 26.69 0.03747

US 12500[Btu/lb] 29.08 0.03439

Source: Platts (2008)

using data provided by Ritschel and Schiffer (2005) that gives lower and upper bounds on average costs for each exporter. We use this information to construct linear average cost curves. The intercept parameterac_x corresponds to the lower bound. To determine the slopebcx we use a second point defined by the maximum production capacity and the upper bound of the average costs assuming a linear average cost function avc_x =ac_x+ bcx·Yx. For the CMT-E model the conversion factorκx must be added to the equations.

Thus, we obtain the following marginal cost functionmcx =κx·ac_x+2·κ²_x·bc_x·Y_x= _∂y^∂cx

xm

that we can integrate in the KKT condition (2.5).

The unit seaborne transport costs trans_cxm expressed in USD/t are based on ob-servations of selected freight rates for each base year 2005 and 2006 and derive from the technical freight literature and IEA (2007a). We use these observations in a regression to obtain linear functions of freight rates based on distance and then to obtain the specific value of trans_c_xm for every possible route between the exporters and the importers.

Finally, the data of production capacity that is available for exports (export mines) is from Kopal (2007). In addition, we include export harbor capacity constraints based on Ritschel and Schiffer (2005) and VDKI (2006). We implicitly assume that shipping (boat) and import harbor capacity is available without capacity limitation.

2.4. Results

Figure 2.1: Imported quantities in the perfect competition (PC), Cournot scenario (CO), and reference data (RE) in 2005, in million tons (Mt) for the CMT model

α_x = 0 for all x; conversely, α_x = 1 for all x in the Cournot scenario. We compare the results of our simulations with the observed trade flows in 2005 and 2006. We also compare the outputs from the two different model specifications to determine the influence of the additional quality information on the results.

Figures 2.1 and 2.2 show the model results and actual trade flows in 2005 for our two specifications CMT and CMT-E. The results of the CMT-E model are converted from Petajoules to million tons using the coal quality data from Table 2.4. Total import quan-tities obtained in both models show a remarkable similarity of the perfect competition results with the reference data for 2005 and 2006. This is also true for 2006 (Figures 2.7 and 2.8 in the Appendix 2.A). The Cournot scenario, on the other hand, gives smaller quantities and considerably higher prices than observed in reality.

The detailed results for both years and both CMT and CMT-E, models are presented in the Appendix 2.A. It can be seen that the number of flows (number of trading relations) in the perfect competition simulation results is small and that most importing countries rely on only one or two suppliers. Real-world flows in 2005 and 2006 showed significantly more diversification of imports. The Cournot results present a more diverse picture with each importer buying from virtually all exporters.

The results of the perfect competition scenario with little diversification are driven by the cost-minimization mechanism that characterizes competitive markets. There is no mark-up on the marginal cost price. Each country imports from the supplier that has the lowest production and transport costs to deliver the coal to that market. In Cournot markets, on the other hand, prices are above their marginal cost levels and attract more

Figure 2.2: Imported quantities in the perfect competition (PC), Cournot scenario (CO), and reference data (RE) in 2005, in million tons (Mt) converted from Petajoules for the CMT-E model

suppliers, including those with higher costs. Although the more diversified trade flow picture makes the Cournot scenario an attractive explanation of the real-world market, we must discard it due to the very high prices and small total quantities obtained when compared to the reference data.

Comparing the 2005 price results of both model specifications with the reference prices (Figure 2.3), we see that the prices of the perfect competition simulations are closer to or in the same range as the real observed prices. However, the relation of the prices between the countries is similar in reality to the Cournot simulation. Cournot competition allows for price discrimination between importing countries whereas the perfect competition simulation does not. This issue will be addressed in Section 2.4.3. These two conclusions also apply to the prices in 2006 (see Figure 2.4). However, the price levels in the model results for 2006 are high (even perfect competition prices are above the real price level) which can be explained by a certain lag in pricing-in capacity constraints (see Section 2.4.2).

When comparing the quantity and energy specification, we observe that the CMT model’s perfect competition results show significantly less supply from Australia than in reality (2005: 66 Mt vs. 110 Mt, 2006: 62 Mt vs. 105 Mt). On the other hand, the CMT model results for supply from Indonesia are higher than in reality by approximatively 10 Mt for both years. One explanation is that a quantity model does not incorporate information about the energy content of the coals. However, in the end, what is im-portant for the customer is the energy contained in the coal. By not incorporating the quality information a distortion is created that makes lower quality coal, like

Indone-2.4. Results

Figure 2.3: CIF Prices in the perfect competition (PC), Cournot scenario (CO), and reference data (RE) in 2005 for the CMT and CMT-E models, USD per ton

Figure 2.4: CIF Prices in the perfect competition (PC), Cournot scenario (CO), and reference data (RE) in 2006 for the CMT and CMT-E models, USD per ton

sian coal, more attractive: the associated transport costs that are mass dependent are underevaluated in comparison to higher quality coals. Reciprocally, the same distortion makes higher quality Australian coal less attractive. This is confirmed by the perfect competition results of the CMT-E model which show higher Australian exports (75 Mt in 2005 and 73 Mt in 2006) and lower Indonesian exports than in the CMT results.

Both model specifications provide evidence that rejects a Cournot market structure for 2005 and 2006. Our perfect competition results of total exported and imported quantities are closer to the real trade volumes. Thus, for the purpose of a market structure analysis a quantity-based model may be sufficient. However, if one requires a better representation of detailed trade flows, a model that incorporates both energy and mass quantity information is needed. This will hold true in the next decades since we expect an increase in the supply of lower quality coal to the world market.

To show the robustness of our results we perform a sensitivity analysis on the price elasticity of the demand function. For this parameter, data sources are scarce and poten-tially outdated. For the base model data set, we chooseεc=−0.3 following Dahl (1993) who reported the short run elasticities for steam coal were between −0.3 and −0.55. Aune et al. (2001) use−0.19for their model. Using these numbers as boundaries for our sensitivity analysis, we run the model with values between −0.2 and−0.6 forε_m. As in the other model runs, we use the same elasticity value for all countriesm.

Figure 2.5 shows that the results of the perfect competition market appear to be more

Figure 2.5: CIF Prices in the perfect competition (PC) and Cournot scenario (CO) model results for different elasticity values, and reference data (RE) in 2005 and 2006 for the CMT-E model, USD per ton

robust than the Cournot results. In 2005, the perfect competition results for the different elasticity values are so close that they can hardly be distinguished visually. Also, the higher the elasticity (in absolute values), the closer the Cournot price results are to the real prices. However, they are still considerably above the observed prices. Given the rather inflexible nature of short-term coal demand due to little switching possibilities to other fuels, and previous elasticity estimations (Dahl, 1993), any value larger than ε_m=−0.6does not seem realistic. Hence, the sensitivity analysis confirms our conclusion that the perfect competition model best represents the international market for steam coal.